Extra Trees is an integrated method composed of many Extra Tree. Its analysis method is very similar to Random Forest, with only two differences. Extra Trees does not use Bagging, that is, when the learner performs regression, it uses the total sample for analysis, and the total sample size and sample composition of each learner are the same. The second difference is that when each learner performs segmentation, Random Forest will calculate the parameter that can reduce the mean square error the most from the feature column and divide it with its threshold value, while the top-down division of Extra Trees is random. It does not calculate the best division point for each feature, but randomly selects a division point from each feature, compares all random division points, and selects the feature and threshold that reduces the mean square error (MSE) the most.

Extra Trees是由許多Extra Tree組成的集成方法，其分析方法與Random Forest非常類似，僅有兩處有差別。Extra Trees並無使用Bagging，即學習器在進行迴歸時，皆使用總樣本進行分析，每一個學習器的總樣本數量和樣本組成皆相同。差別二則是每一個學習器再進行分割時，Random Forest會從特徵列中計算能使均方誤差下降最多的參數與其閥值進行分割，而Extra Trees自上而下的劃分是隨機的，它並不計算每個特徵的最佳劃分點，而是從每個特徵中各隨機選擇一個劃分點，比較所有隨機的劃分點後選擇使均方誤差(MSE)下降最多的特徵與閥值進行分割。